Particle size distribution and morphology are among the factors that affect the physical properties of particulate systems (e.g. powder mixtures, pastes, castable refractories). One such property is packing density. There are, in the literature, models that optimise particle size distributions for maximum packing density, all of them derived for spherical particles (e.g. Furnas, Andreasen, Alfred). In this work, commercial alumina powders (reactive and tabular) were divided into nine particle size classes. Following two different approaches, the latter were used to build six different ternary systems of complementary particle sizes. Using the response surface methodology and related statistical techniques (software Statistica), the particle size distribution that maximises the packing density was obtained in both cases and, by comparison with theoretical particle size distributions, the validity of Alfred's theoretical model for perfect spheres was demonstrated. These results clearly show that the harmful effect of the non-spherical shape of real particles can, in fact, be compensated by the optimization of the overall particle size distribution.
particle packing; particle size distributions; statistical analysis
